## Application of a new algorithm in the differential diagnosis of wide QRS complex tachycardia

## Abstract

**Aims** The Brugada criteria proposed to distinguish between regular, monomorphic wide QRS complex tachycardias (WCT) caused by supraventricular (SVT) and ventricular tachycardia (VT) have been reported to have a better sensitivity and specificity than the traditional criteria. By incorporating two new criteria, a new, simplified algorithm was devised and compared with the Brugada criteria.

**Methods and results** A total of 453 WCTs (331 VTs, 105 SVTs, 17 pre-excited tachycardias) from 287 consecutive patients with a proven electrophysiological (EP) diagnosis were prospectively analysed by two of the authors blinded to the EP diagnosis. The following criteria were analysed: (i) presence of AV dissociation; (ii) presence of an initial *R* wave in lead aVR; (iii) whether the morphology of the WCT correspond to bundle branch or fascicular block; (iv) estimation of initial (*v*_{i}) and terminal (*v*_{t}) ventricular activation velocity ratio (*v*_{i}/*v*_{t}) by measuring the voltage change on the ECG tracing during the initial 40 ms (*v*_{i}) and the terminal 40 ms (*v*_{t}) of the same bi- or multiphasic QRS complex. A *v*_{i}/*v*_{t} >1 was suggestive of SVT and a *v*_{i}/*v*_{t} ≤1 of VT. An initial *R* wave in lead aVR suggested VT. The overall test accuracy of the new algorithm was superior (*P* = 0.006) to that of the Brugada criteria. The new algorithm had a greater sensitivity (*P* < 0.001) and (−) predictive value (NPV) for VT diagnosis, and specificity (*P* = 0.0471) and (+) predictive value (PPV) for SVT diagnosis than those of the Brugada criteria [both NPV for VT diagnosis and PPV for SVT diagnosis were: 83.5% (95% confidence interval = CI 75.9–91.1%) for the new vs. 65.2% (95% CI 56.5–73.9%) for the Brugada algorithms].

**Conclusion** The new algorithm is a highly accurate tool for correctly diagnosing the cause of WCT ECGs.

- Wide QRS complex tachycardia
- Brugada criteria
- Ventricular tachycardia
- Supraventricular tachycardia

**See page 525 for the editorial comment on this article (doi:10.1093/eurheartj/ehl557)**

## Introduction

Wide QRS complex tachycardia (WCT) is a common arrhythmia with important therapeutic and prognostic implications and often presents a diagnostic challenge. WCTs may be ventricular in origin or may be supraventricular, conducted with fixed or functional bundle branch block (BBB) pattern, or supraventricular due to drug or electrolyte-induced changes or pre-excitation. Pre-excited tachycardias (PXT) and drug- and electrolyte-induced WCTs account for only a small minority (1–5%) of causes of WCT. Because most WCTs are either ventricular tachycardia (VT) or supraventricular tachycardia (SVT), conducted with fixed or functional BBB pattern, the clinically relevant problem in the differential diagnosis of WCTs is the differentiation of the latter two.^{1} The ECG remains the cornerstone of distinguishing SVT from VT. A bewildering number of ECG criteria have been reported^{2–17} for the differential diagnosis of regular WCTs. Using all these traditional ECG criteria, an accurate diagnosis is now possible in about 90% of WCTs.^{2,3,10} However, many of these criteria are complicated and not consistently present, thus not useful in an urgent setting. Brugada *et al*.^{7} proposed a relatively simple, stepwise, decision tree-like algorithm to differentiate between WCTs due to VT and SVT. However, that algorithm still retained the traditional morphological criteria in its last step. They reported that this algorithm had a sensitivity (98.7%) and specificity (96.5%) superior to those of the currently available criteria. Other authors^{1,10,18} also found the Brugada criteria useful, though reported a lower sensitivity and specificity. Our aim was to devise another simplified, new algorithm for the differential diagnosis of WCTs by eliminating most of the complicated morphological criteria and compare it with the Brugada criteria.

## Methods

A different set of patients was used to devise the algorithm from that used to test the already established algorithm. We used retrospectively 103 WCTs available in the database of Indiana University obtained from patients with proven electrophysiological (EP) diagnosis referred to EP study either because of spontaneous WCT or because of other clinical reasons and WCT was induced during the EP study. Then, to test the established algorithm, 453 regular WCT (331 VTs, 105 SVTs, 17 PXTs) tracings recorded from 287 consecutive patients during EP study conducted from June 1998 to November 2004 at Indiana University with proven EP diagnosis were prospectively analysed by two of the authors blinded to the EP diagnosis and the patients' clinical data. An informed consent exemption was obtained from the Indiana University Institutional Review Board for analysis of a deidentified dataset. The observers were given complete 12-lead standard ECGs obtained during tachycardia for analysis. WCT was defined as a rhythm with a rate ≥100 b.p.m. with a QRS duration ≥120 ms. Only monomorphic WCTs were analysed using the following criteria: (i) presence of A–V dissociation; (ii) presence of an initial *R* wave in lead aVR; (iii) whether the morphology of the WCT correspond to BBB or fascicular block [the diagnostic criteria proposed by Willems *et al*.^{19} for intraventricular conduction disturbances were used (see *Table 1*)]; (iv) an index of slow conduction at the beginning and at the end of the QRS complex by estimation of initial (*v*_{i}) and terminal (*v*_{t}) ventricular activation velocity ratio (*v*_{i}/*v*_{t}), obtained by measuring the voltage in millivolts on the ECG tracing the impulse travelled vertically during the initial 40 ms (*v*_{i}) and the terminal 40 ms (*v*_{t}) of the same bi- or multiphasic QRS complex. The A–V dissociation criterion is identical in both algorithms (first criterion of the new and third of the Brugada algorithms). The *v*_{i} and *v*_{t} were measured in an individual QRS complex in any lead having a bi- or multiphasic QRS complex, in which the onset and end of the QRS were clearly visible and the initial ventricular activation was the most rapid (fastest). When either the initial or terminal 40 ms of the QRS complex displayed both positive and negative deflections, the sum of their absolute values (disregarding polarity) were used as the values of *v*_{i} and *v*_{t}. Because three channels were recorded simultaneously on the ECG tracings, the onset and end of the QRS were defined by the earliest and latest ventricular depolarization, respectively, among the three simultaneously recorded leads that included the lead with the fastest initial ventricular activation. Most frequently (in 87% of WCTs), the *v*_{i} was the fastest in the precordial leads and the leads most commonly used for analysis of *v*_{i}/*v*_{t} were *v*_{3}, *v*_{5}, and *v*_{2} in decreasing order of frequency. Thus, limb leads were used to determine *v*_{i}/*v*_{t} in only 13% of WCTs. We hypothesized that a *v*_{i}/*v*_{t} >1 was suggestive of SVT and a *v*_{i}/*v*_{t} ≤1 of VT (*Figures 1* and *2*). The *v*_{i}/*v*_{t} criterion was validated in 111 ECG tracings recorded during sinus rhythm in patients with all types of intraventricular conduction disturbances, some of whom also had old myocardial infarction (MI). The *v*_{i}/*v*_{t} was >1 (signifying supraventricular origin) in 22/25 (88%) tracings with left BBB pattern, in 55/56 (98%) with right BBB pattern, and 27/30 (90%) with non-specific intraventricular block pattern. The presence of an initial *R* wave (such as *R* or *RS* wave, but not *rS* wave) in lead aVR was hypothesized to suggest VT. The four criteria of the new algorithm were organized in a stepwise, decision-tree format similar to that of the Brugada algorithm (*Figure 3*). The four steps were used in the following sequence. (1) If A–V dissociation was present, the diagnosis of VT was made and the analysis was stopped. (2) If an initial *R* wave was present in lead aVR, the diagnosis of VT was made and the analysis was stopped. (3) If the morphology of WCT did not correspond to BBB or fascicular block, the diagnosis of VT was made and the analysis was stopped. (4) In the last step, when the *v*_{i}/*v*_{t} was ≤1, the diagnosis of VT was made, and if the *v*_{i}/*v*_{t} was >1, the diagnosis of SVT. Our algorithm, as well as the traditional morphological ECG criteria, are unable to reliably differentiate VTs from PXTs in most WCT cases (with the exception of the presence of A–V dissociation and possibly that of an initial *R* wave in lead aVR along with other criteria suggested by Antunes *et al*.^{9} that are infrequently present) and thus, the final diagnosis of VT in the third and fourth steps of the algorithm included also PXTs. *Figure 4* demonstrates an example how the new algorithm was applied.

Definition of ventricular conduction delays | |
---|---|

A. Complete BBBs | B. Fascicular blocks |

Qualifying statements | 1. Left anterior fascicular block (LAFB) |

S1) QRS duration ≥0.12 s (adults) | Qualifying statements |

S2) Supraventricular rhythm | S1) QRS duration <0.12 s |

S3) Absence of WPW pattern | S2) QRS axis ≤−30° |

Criteria for a complete BBB | S3) rS pattern in II and III and aVF |

a) S1 and S2 and S3 | S4) qR pattern in aVL |

1. Complete right BBB (RBBB) | S5) R peak time ≥0.045 s in aVL |

Qualifying statements | S6) Slurred R downstroke in aVL |

S1) R′ or r′ in V_{1} or V_{2} | S7) Slurred S in V_{5} or V_{6} |

S2) S duration > R duration in I and V_{6} | Criteria for uncomplicated LAFB |

S3) S duration >0.04 s in I and V_{6} | a) S1 and S2 and S3 and S4 and S5 or |

S4) R peak time >0.05 s in V_{1} or V_{2} | b) S1 and S2 and S3 and S4 and S6 or |

Criteria for RBBB | c) S1 and S2 and S3 and S4 and S7 |

a) S1 and S2 orb) S1 and S3 orc) S4 and (S2 or S3) | Qualifying statement S3 is usually present with criteria a, b, and c above. If there is a QS in lead II, LAFB cannot be differentiated from inferior MI. |

2. Complete left BBB (LBBB) | 2. Left posterior fascicular block (LPFB) |

Qualifying statements | Qualifying statements |

S1) Broad and notched or slurred R in I and V_{5} or V_{6} | S1) QRS duration <0.12 s |

S2) Absence of Q wave in I and V_{5} or V_{6} | S2) 180° > QRS axis >90° |

S3) R peak times ≥0.06 s in V_{5} or V_{6} | S3) qR pattern in III and aVF with Q duration ≤0.04 sec |

S4) Absence of other causes of right axis deviation | |

Criteria for LBBB | Criteria for LPFB |

a) S1 and S2 and S3 | a) S1 and S2 and S3 and S4 |

3. Non-specific intraventricular block All cases with QRS duration >0.12 s which do not meet the criteria for LBBB or RBBB |

From Willems

*et al*.^{19}with modifications. Definitions for incomplete BBBs were omitted, because the QRS duration of the WCT tracings analysed in this study was ≥0.12 s.

### Statistical analysis

Occurrence of true positive and negative, false positive and negative results expressed as percentage of the total number of observations as well as sensitivity and specificity were compared between two algorithms by first constructing 2 × 2 cross tables demonstrating where the two algorithms agreed or disagreed and then by using the non-parametric McNemar's test for comparing two related proportions, to determine which algorithm was better. The SPSS 13 for Windows software package (SPSS Inc., Chicago, IL, USA) was used for statistical analysis. A *P* < 0.05 value was considered statistically significant. However, the above described method was not suitable for the comparison of the predictive values, because in this case the denominators for the two algorithms differ (unlike specificity and sensitivity, where the denominators are the same). Lacking an entirely appropriate statistical method to compare the predictive values, these are presented simply with 95% confidence intervals (CI) without statistical comparison, and a significant between-groups difference in algorithms is indicated by disjoint (non-overlapping) CIs. Some patients are in the dataset more than once (several VTs with different morphology were induced in some patients while a few had WCTs due to both SVT and VT, occurred during the same EP study). Because these episodes behaved as independent, unrelated events, they were analysed as different WCT tracings in the study.

The Kappa statistic was used to quantify overall interobserver agreement using SAS statistical software package (SAS/STAT Software Release 6.12, SAS Institute Inc., Cary, NC, USA). Overall interobserver agreement was defined as good if κ>0.6, moderate if 0.6>κ>0.4, and poor if κ<0.4.

## Results

### Patient characteristics

The patient groups differed in that the PXT and SVT groups had younger patients, more females, and fewer patients with a history of prior MI or cardiomyopathy and far more patients without structural heart disease than the VT group (*Table 2*). No patient in the PXT group and fewer patients in the SVT group took antiarrhythmic drugs or had pre-existent BBB than in the VT group.

SVT n = 105 | VT n = 331 | PXT n = 17 | |
---|---|---|---|

Age (years) (mean ± SD) | 45 ± 20 | 57 ± 17 | 36 ± 17 |

Female/male (%) | 44/56 | 17/83 | 31/69 |

Antiarrhythmic drugs (%) | 4 | 45 | 0 |

Pre-existent BBB(%) | 25 | 35 | 0 |

Past history | |||

Post-MI (%) | 4 | 61 | 0 |

Cardiomyopathy (%) | 1 | 15 | 0 |

No structural heart disease (idiopathic) (%) | 93 | 11 | 100 |

### Overall test accuracy

The new algorithm correctly classified 409 of 453 WCTs [90.3% (95% CI 87.6–93%) overall test accuracy (TA)] and was superior (*P* = 0.006) to that of Brugada algorithm [384/453 (84.8% overall TA) (95% CI 81.5–88.1%)] (*Table 3*). *Figure 5* shows a WCT misclassified by the Brugada criteria and classified correctly using the new algorithm. In the first step, the A–V dissociation criterion correctly diagnosed VT in 100%, in the second step, the aVR criterion in 97.6%, and in the third step, the BBB or fascicular block criterion in 89.1% of cases. In the fourth step, the *v*_{i}/*v*_{t} criterion correctly classified 111/135 [82.2% (95% CI 75.8–88.7%)] WCTs and was applicable in all cases. The diagnostic accuracy of each criterion was also evaluated individually in all WCT tracings. The TA of each criterion calculated in both ways was similar (*Table 3*). The TA of the first and second Brugada criteria were also good (>90%), however, that of the fourth Brugada criterion was significantly lower [68% (95% CI 60.5–75.6%) vs. 82.2% (95% CI 75.8–88.7%), *P* = 0.004] than the TA of the *v*_{i}/*v*_{t} criterion in the fourth step. Among all ECGs, the *v*_{i}/*v*_{t} criterion could not be applied in 16/453 (3.5%) cases, either because no bi- or multiphasic QRS complex was found in any of the 12 ECG leads or in some fast WCTs when the onset or end of the QRS complex could not be discerned. *Figures 4* and *6* demonstrate examples for difficulties in the determination of the QRS onset and end. The *v*_{i}/*v*_{t} criterion was thus applicable in 437/453 (96.5%) of WCTs; its overall TA was 359/437 [82.2% (95% CI 78.6–85.7%)], similar to that of the Brugada criteria [84.8% (95% CI 81.5–88.1%)] and inferior (*P* < 0.001) to a combination of all criteria of the new algorithm [90.3% (95% CI 87.6–93%)]. Interestingly, all 16 WCTs where the *v*_{i}/*v*_{t} criterion could not be applied were VTs. A total of 18 WCT episodes were misclassified by both the new and Brugada algorithms (*Figure 6*). The two observers produced very similar results: the interobserver variability was non-significant, as was the difference between the number of misclassified ECGs using both algorithms (results not shown). Therefore, only the results from observer 1 are published and used for analysis. *Figure 7* demonstrates the numbers of VT and SVT, true and false positive diagnoses made in each step of the new algorithm.

Criterion | Correct diagnosis |
---|---|

All four criteria of the new algorithm | 409/453 [90.3 (87.6–93)] |

A–V dissociation = third Brugada | 35/35 [100 (100–100)] |

aVR (in all ECGs) | 135/138 [97.8 (95.4–100.3)] |

aVR (in the second step) | 124/127 [97.6 (95–100.3)] |

BBB, FB (in all ECGs) | 273/293 [93.2 (90.3–96.1)] |

BBB, FB (in the third step) | 139/156 [89.1 (84.2–94)] |

v_{i}/v_{t} (in all ECGs) | 359/437 [82.2** (78.6–85.7)] |

v_{i}/v_{t} (in the fourth step) | 111/135 [82.2 (75.8–88.7)] |

First Brugada | 79/85 [92.9 (87.5–98.4)] |

Second Brugada | 195/212 [92.5 (88.3–95.6)] |

Fourth Brugada | 100/147 [68*** (60.5–75.6)] |

All Brugada | 384/453 [84.8* (81.5–88.1)] |

The numbers represent the correct diagnoses/total number of tracings investigated with the criterion [percentage = TA (95% CI)]. The overall (both for VT and SVT diagnoses) TA of all four criteria of the new algorithm, all Brugada criteria and the

*v*_{i}/*v*_{t}criterion applied to all ECGs were compared statistically. Also the overall TA of the*v*_{i}/*v*_{t}criterion applied in the fourth step was compared with that of the fourth Brugada criterion separately.*

*P*< 0.01, ***P*< 0.001 vs. all criteria of the new algorithm; ****P*< 0.01 for the fourth Brugada criterion vs. the*v*_{i}/*v*_{t}criterion applied in the fourth step.

### Sensitivity, specificity, and predictive values in VT diagnosis

Because only two final diagnoses (VT or SVT) were possible with the algorithms used, the specificity and positive predictive value (PPV) for VT diagnosis were the same as the sensitivity and negative predictive value (NPV) for SVT diagnosis (respectively), and inversely, the sensitivity and NPV for VT diagnosis were the same as the specificity and PPV for SVT diagnosis, respectively. The sensitivity [95.7% (95% CI 93.6–97.8%) vs. 88.2% (95% CI 84.8–91.6%), *P* < 0.001] and NPV [83.5% (95% CI 75.9–91.1%) vs. 65.3% (95% CI 56.7–73.8%)] for VT diagnosis of the new algorithm were superior to those of the Brugada criteria (*Table 4*). The specificity of the *v*_{i}/*v*_{t} criterion applied to all ECGs was greater for VT diagnosis than that of a combination of all criteria of the new algorithm and that of the Brugada algorithm [81.9% (95% CI 74.5–89.3%) vs. 72.4% (95% CI 63.8–80.9%) and 73.3% (95% CI 64.9–81.8%) respectively, *P* = 0.004 for the new and *P* = 0.0173 for the Brugada algorithm]. However, the sensitivity [95.7% (95% CI 93.6–97.8%) vs. 82.2% (95% CI 78.1–86.3%), *P* < 0.001] and NPV [83.5% (95% CI 75.9–91.1%) vs. 59.7% (95% CI 51.7–67.7%)] of the combination of all criteria of the new algorithm and the sensitivity of the Brugada algorithm [88.2% (95% CI 84.8–91.6%) vs. 82.2% (95% CI 78.1–86.3%), *P* = 0.0277] were superior to those of the *v*_{i}/*v*_{t} criterion alone applied to all ECGs. Among the other individual criteria, only the BBB, fascicular block criterion had a fairly high sensitivity (74.7%); all other individual criteria had relatively low sensitivity, despite their good specificity (*Table 4*). The aVR criterion was never positive in the 17 WCT episodes due to PXT, suggesting that it may be useful not only for distinguishing VT from SVT but also VT from PXT. The *v*_{i}/*v*_{t} criterion applied in the fourth step had a significantly greater sensitivity [70% (95% CI 57.3–82.7%) vs. 39.4% (95% CI 27.6–51.2%), *P* < 0.001] and NPV [83.5% (95%CI 75.9–91.1%) vs. 65.2% (95% CI 56.5–73.9%)] than the fourth Brugada criterion for VT diagnosis.

Criterion | Sensitivity | Specificity | (+) Predictive value | (−) Predictive value | ||||
---|---|---|---|---|---|---|---|---|

VT Dx | SVT Dx | VT Dx | SVT Dx | VT Dx | SVT Dx | VT Dx | SVT Dx | |

All new algorithm criteria | 95.7 (93.6–97.8) | 72.4 (63.8–80.9) | 72.4 (63.8–80.9) | 95.7 (93.6–97.8) | 92 (89.2–94.8) | 83.5 (75.9–91.1) | 83.5 (75.9–91.1) | 92 (89.2–94.8) |

A–V dissociation | 10.1 (6.9–13.2) | 100 (100–100) | 100 (100–100) | 25.1 (21–29.3) | ||||

aVR (in the second step) | 39.6 (34.2–45) | 97.1 (94–100.3) | 97.6 (95–100.3) | 35.1 (29.6–40.5) | ||||

BBB, FB (in the third step) | 74.7 (68.5–81) | 83.3 (76.1–90.6) | 89.1 (84.2–94) | 64.4 (56.2–72.6) | ||||

v_{i}/v_{t} (in the 4^{th} step) | 70 (57.3–82.7) | 89.4 (82.9–96) | 89.4 (82.9–96) | 70 (57.3–82.7) | 79.5 (67.6–91.5) | 83.5 (75.9–91.1) | 83.5 (75.9–91.1) | 79.5 (67.6–91.5) |

v_{i}/v_{t} (in all ECGs) | 82.2*^{,†} (78.1–86.3) | 81.9^{†} (74.5–89.3) | 81.9**^{,†} (74.5–89.3) | 82.2** (78.1–86.3) | 93.2 (90.3–96.1) | 59.7^{a} (51.7–67.7) | 59.7^{a} (51.7–67.7) | 93.2 (90.3–96.1) |

First Brugada | 22.8 (18.4–27.2) | 94.3 (89.9–98.7) | 92.9 (87.5–98.4) | 27.2 (22.6–31.7) | ||||

Second Brugada | 56.5 (51.3–61.7) | 84.9 (78.1–91.7) | 92.5 (88.9–96) | 37.3 (31.2–43.5) | ||||

Fourth Brugada | 39.4^{δδδ} (27.6–51.2) | 91.5 (85.4–97.5) | 91.5 (85.4–97.5) | 39.4^{δδδ} (27.6–51.2) | 78.8 (64.8–92.7) | 65.2^{b} (56.5–73.9) | 65.2^{b} (56.5–73.9) | 78.8 (64.8–92.7) |

All Brugada | 88.2* (84.8–91.6) | 73.3 (64.9–81.8) | 73.3 (64.9–81.8) | 88.2*** (84.8–91.6) | 91.6 (88.7–94.6) | 65.3^{a} (56.7–73.8) | 65.3^{a} (56.7–73.8) | 91.6 (88.7–94.6) |

Dx, diagnosis. Bold numbers represent percentage values; italic numbers in parentheses are 95% CI. The sensitivity, specificity, and predictive values of all four criteria of the new algorithm, all Brugada criteria, and the

*v*_{i}/*v*_{t}criterion applied to all ECGs were compared statistically and those of the*v*_{i}/*v*_{t}criterion applied in the fourth step were also compared with those of the fourth Brugada criterion separately. A significant between-groups difference in predictive values (^{a}vs. all criteria of the new algorithm,^{b}fourth Brugada criterion vs. the*v*_{i}/*v*_{t}criterion applied in the fourth step) is indicated only by disjoint (non-overlapping) 95% CI. For sensitivity and specificity: **P*< 0.001, ***P*< 0.01, ****P*< 0.05, vs. all criteria of the new algorithm,^{†}significant. (^{†}*P*< 0.05,^{††}*P*< 0.01,^{†††}*P*< 0.001) difference between the*v*_{i}/*v*_{t}criterion in all ECGs and all Brugada criteria,^{§}*P*< 0.05,^{§§}*P*< 0.01,^{δδδ}*P*< 0.001 for the fourth Brugada criterion vs. the*v*_{i}/*v*_{t}criterion applied in the fourth step.

### Sensitivity, specificity, and predictive values in SVT diagnosis

The specificity [95.7% (95% CI 93.6–97.8%) vs. 88.2% (95% CI 84.8–91.6%), *P* = 0.0471] and PPV [83.5% (95% CI 75.9–91.1%) vs. 65.3% (95% CI 56.7–73.8%)] for SVT diagnosis of the new algorithm were superior to those of the Brugada algorithm (*Table 4*). The specificity [95.7% (95% CI 93.6–97.8%) vs. 82.2% (95% CI 78.1–86.3%), *P* < 0.01] and PPV [83.5% (95% CI 75.9–91.1%) vs. 59.7% (95% CI 51.7–67.7%)] of the combination of all criteria of the new algorithm proved to be superior to those of the *v*_{i}/*v*_{t} criterion alone applied to all ECGs. The *v*_{i}/*v*_{t} criterion applied in the fourth step had a significantly greater specificity [70% (95% CI 57.3–82.7%) vs. 39.4% (95% CI 27.6% to 51,2%), *P* < 0.001] and PPV [83.5% (95% CI 75.9–91.1%) vs. 65.2% (95% CI 56.5–73.9%)] than the fourth Brugada criterion for SVT diagnosis.

### Subgroup analysis

In the presence of pre-existent BBB, the overall TA of the new algorithm was superior [(92.2% (95% CI 87.8–96.6%) vs. 85.8% (95% CI 80.1–91.6%), *P* = 0.027] and in the presence of idiopathic VT was borderline superior [(86.5% (95% CI 75.5–97.5%) vs. 67.6% (95% CI 52.5–82.7%), *P* = 0.065] to that of the Brugada algorithm (*Table 5*). When both pre-existent BBB and class I antiarrhythmic drug or amiodarone treatment were present, the overall TA of the new algorithm was borderline superior [97.6% (95% CI 94.4–100.9%) vs. 92.9% (95% CI 87.3–98.4%), *P* = 0.063] to that of the Brugada criteria. For VT diagnosis, the new algorithm had a significantly better sensitivity [100% (95% CI 100–100%) vs. 95.9% (95% CI 92.7–99.1%), *P* = 0.031] and NPV [100% (95% CI 100–100%) vs. 33.3% (95% CI 2.5–64.1%)] in the presence of class I antiarrhythmic drug or amiodarone treatment, and in the presence of pre-existent BBB, a borderline superior sensitivity [98.3% (95%CI 95.9–100.7%) vs. 92.2% (95% CI 87.3–97.1%), *P* = 0.065], and a borderline superior NPV [89.5% (95% CI 75.7–103.3%) vs. 62.5% (95% CI 43.1–81.9%)] compared with the Brugada criteria. The sensitivity of the new criteria in diagnosing VT in the case of idiopathic VTs was borderline superior [86.5% (95% CI 75.5–97.5%) vs. 67.6% (95% CI 52.5–82.7%), *P* = 0.065] to that of the Brugada criteria. The PPV for SVT diagnosis of the new criteria was borderline superior [89.5% (95% CI 75.7–103.3%) vs. 62.5% (95% CI 43.1–81.9%)] to that of the Brugada criteria in the presence of pre-existent BBB.

Class I antiarrhythmic drugs or amiodarone n = 151 | Pre-existent BBB n = 141 | Class I antiarrhythmic drugs or amiodarone + pre-existent BBB n = 84 | Idiopathic VT n = 37 | ||||||
---|---|---|---|---|---|---|---|---|---|

New algorithm | Brugada criteria | New algorithm | Brugada criteria | New algorithm | Brugada criteria | New algorithm | Brugada criteria | ||

Overall TA | % (95% CI) | 98.7 (96.9–100.5) | 95.4 (92–98.7) | 92.2* (87.8–96.6) | 85.8 (80.1–91.6) | 97.6** (P = 0.063) (94.4–100.9) | 92.9 (87.3–98.4) | 86.5** (P = 0.065) (75.5–97.5) | 67.6 (52.5–82.7) |

Sensitivity for VT Dx | % (95% CI) | 100* (100–100) | 95.9 (92.7–99.1) | 98.3** (P = 0.065) (95.9–100.7) | 92.2 (87.3–97.1) | 100* (100–100) | 93.8 (88.4–99.1) | 86.5** (P = 0.065) (75.5–97.5) | 67.6 (52.5–82.7) |

Specificity for VT Dx | % (95% CI) | 50 (1–99) | 75 (32.6–117.4) | 65.4 (47.1–83.7) | 57.7 (38.7–76.7) | 50 (1–99) | 75 (32.6–117.4) | ||

(+) pred. val. for VT Dx | % (95% CI) | 98.7 (96.8–100.5) | 99.3 (97.9–100.7) | 92.6 (88–97.3) | 90.6 (85.3–95.9) | 97.6 (94.2–100.9) | 98.7 (96.1–101.2) | 100 (100–100) | 100 (100–100) |

(−) pred. val. for VT Dx | % (95% CI) | 100*** (100–100) | 33.3 (2.5–64.1) | 89.5 (75.7–103.3) | 62.5 (43.1–81.9) | 100 (100–100) | 37.5 (4–71) | ||

Sensitivity for SVT Dx | % (95% CI) | 50 (1–99) | 75 (32.6–117.4) | 65.4 (47.1–83.7) | 57.7 (38.7–76.7) | 50 (1–99) | 75 (32.6–117.4) | ||

Specificity for SVT Dx | % (95% CI) | 100 (100–100) | 95.9 (92.7–99.1) | 98.3 (95.9–100.7) (87.3–97.1) | 92.2 (100–100) | 100 (88.4–99.1) | 93.8 | ||

(+) pred. val. for SVT Dx | % (95% CI) | 100 (100–100) | 33.3 (2.5–64.1) | 89.5 (75.7–103.3) | 62.5 (43.1–81.9) | 100 (100–100) | 37.5 (4–71) | ||

(−) pred. val. for SVT Dx | % (95% CI) | 98.7 (96.8–100.5) | 99.3 (97.9–100.7) | 92.6 (88–97.3) | 90.6 (85.3–95.9) | 97.6 (94.2–100.9) | 98.7 (96.1–101.2) |

Dx, diagnosis; pred. val., predictive value, 95% CI = 95% CI, *Significant (

*P*< 0.05), **borderline significant difference between the new and Brugada algorithms. ***Significant difference in predictive values between the new and Brugada algorithms indicated by disjoint (non-overlapping) 95% CI. The missing values in the idiopathic VT column either could not be calculated, because the number of true negatives in VT diagnosis and that of true positives in SVT diagnosis were zeroes, or was no sense to calculate them (SVT diagnosis parameters).

## Discussion

### Major findings

Our new algorithm for the differential diagnosis of WCTs has been shown to have a significantly better overall TA: a greater sensitivity and NPV in VT diagnosis and a greater PPV and specificity in SVT diagnosis compared with the Brugada criteria. The overall TA of our relatively simple new algorithm, which eliminated most of the difficult-to-recall morphological criteria, was on a par with the use of all published traditional ECG criteria.^{3,10}

### The rationale behind the two new criteria and their potential value

The rationale behind the *v*_{i}/*v*_{t} criterion is that during WCT due to SVT, the initial activation of the septum should be invariably rapid and the intraventicular conduction delay causing the wide QRS complex occurs in the mid to terminal part of the QRS. Thus, the conduction velocity of initial ventricular activation should be faster than that of the later or terminal ventricular activation during SVT conducted with functional aberration or fixed BBB. During WCT due to VT, however, an initial slower muscle-to-muscle spread of activation occurs until the impulse reaches the His–Purkinje system, after which the rest of the ventricular muscle is more rapidly activated. Thus, in WCTs due to VT, the conduction velocity of initial ventricular activation is slower than that of the later ventricular activation. This assumption should hold true regardless of the mechanism of VT or presence or absence of structural heart disease. We used another assumption while devising the *v*_{i}/*v*_{t} criterion, that the steepness of the QRS (which was measured by voltage in millivolts, the impulse travelled in vertical direction during a given time period) is directly proportional with the conduction velocity of the propagating impulse in the ventricle. Antiarrhythmic drugs that impair conduction in the His–Purkinje system and/or ventricular myocardium (such as class I drugs and amiodarone) would be expected to decrease the *v*_{i} and *v*_{t} approximately to the same degree, therefore the *v*_{i}/*v*_{t} ratio will not change significantly. The reasons for misdiagnoses using the *v*_{i}/*v*_{t} criterion alone might be: (i) disorders involving the myocardium locally can alter the *v*_{i} or *v*_{t}, for example, a decreased *v*_{i} with unchanged *v*_{t} may be present in the case of an SVT occurring in the presence of an anteroseptal MI leading to the misdiagnosis of VT; or a scar situated at a late activated ventricular site may result in a decreased *v*_{t} in the presence of VT leading to the misdiagnosis of SVT; (ii) in the case of a fascicular VT, the *v*_{i} is not slower than the *v*_{t}; (iii) if the exit site of the re-entry circuit is very close to the His–Purkinje system, it might result in a VT with a relatively narrow QRS complex and the slowing of the *v*_{i} may last for such a short time that it cannot be detected by the surface ECG.

Because all 16 WCTs where the *v*_{i}/*v*_{t} criterion could not be applied proved to be VTs, the mere fact that the *v*_{i}/*v*_{t} criterion cannot be applied might indicate that the underlying mechanism of WCT is VT. However, this observation needs further confirmation.

The aVR criterion is not completely new in the sense that it is similar to the old QRS axis criterion, according to which the QRS axis in the right superior quadrant (−90° to ± 180°) suggests VT, because the resultant QRS vector should be −60° to +120° to give rise a predominantly positive QRS in lead aVR. However, the aVR criterion is different from the old QRS axis criterion not only in the minimal difference in QRS axis that is needed to have an *R* wave in lead aVR vs. a right superior quadrant axis, but also in the fact that our aVR criterion suggests VT only in the presence of an initial *R* wave in lead aVR. *Figure 8* demonstrates why the aVR criterion is different and may be superior to the QRS axis criterion. A WCT due to SVT is shown with a predominantly positive QRS complex in lead aVR, the QRS axis in the frontal plane is −160°, thus the QRS axis criterion suggests VT. However, the onset of the predominantly positive QRS in lead aVR is negative (*Q* wave), therefore the aVR criterion did not suggest VT. A positive aVR criterion suggesting VT seems to exclude PXT: none of our 17 PXTs had a positive aVR criterion. However, this potential use of aVR criterion needs further testing. The observation that an initial *R* wave in lead aVR rules out PXT is consistent with the fact that activation of the ventricles over an accessory pathway proceeds from the base towards the apex of the heart yielding a negative QRS complex in lead aVR. An initial *R* wave may be present in lead aVR, resulting in an rS complex, as a normal variant or in the presence of inferior MI due to loss of initial inferiorly directed forces and reciprocal gain in unopposed superiorly directed forces. However, in normal sinus rhythm the R/S ratio in lead aVR should be <1 thus, an initial *R* wave should not be present.^{20,21}

### Possible explanation for the superiority of the new algorithm to the Brugada criteria

The fourth step of the Brugada algorithm involving the complicated morphological ECG criteria accounted for most [41/70(59%)] of incorrectly diagnosed WCT episodes. Although in the first three steps of the two algorithms the TAs were quite similar, the *v*_{i}/*v*_{t} criterion in the fourth step proved to be superior to the fourth Brugada criterion (having a significantly greater TA, sensitivity, and NPV for VT diagnosis, and specificity and PPV for SVT diagnosis). Another potential cause for the lower overall TA of the Brugada algorithm may be that it uses highly specific but relatively insensitive criteria (in this study the sensitivity for VT diagnosis of the four Brugada criteria in the order of their application was 22.8, 56.5, 10.1, and 39.4%, respectively) and the criteria of the new algorithm are not only highly specific but some of them have a good sensitivity (the sensitivity of the BBB, fascicular block criterion was 74.7% and that of the *v*_{i}/*v*_{t} criterion in the fourth step was 70% in VT diagnosis). Furthermore, it was shown^{18} that the presence of pre-existing BBB and the use of class I antiarrhythmic drugs or amiodarone result in a low specificity of the second Brugada criterion (i.e. in many of these patients with WCT due to SVT, the longest R-to-S interval in the precordial leads will be >100 ms suggesting the misdiagnosis of VT). However, pre-existing BBB and class I drug or amiodarone treatment are not expected to influence the *v*_{i}/*v*_{t} criterion or any other criteria of the new algorithm.^{1} Indeed, in 8/30 (27%) SVTs that were present in patients taking either class I antiarrhythmic drugs or amiodarone or having pre-existent BBB, the longest R-to-S interval was >100 ms in our study. Also consistent with this finding, our results showed that in the presence of pre-existent BBB the overall TA of the new algorithm was superior to that of the Brugada criteria. Another weak point of the RS >100 ms Brugada criterion is that the RS interval during idiopathic intrafascicular tachycardia is 60 to 80 ms, therefore these VTs cannot be correctly diagnosed using the second Brugada criterion.^{22–24} Among all idiopathic VTs in our study, the longest R-to-S interval was <100 ms in 15/29 (52%) cases. As noted previously, the new algorithm demonstrated a borderline superior overall TA compared with that of the Brugada criteria in the presence of idiopathic VT.

### Limitations

The new algorithm is inherently unable to recognize certain forms of WCT. Bundle branch re-entry VT, fascicular VT, and SVT involving an atriofascicular accessory pathway are associated with typical BBB pattern indistinguishable from that associated with SVT with functional aberrancy or pre-existent BBB,^{1,2,10,25} unless A–V dissociation is present. Another limitation of the new algorithm is the somewhat arbitrary definition that *v*_{i}/*v*_{t} should be measured in the lead where initial ventricular activation is the fastest. The underlying premise was that, in VT, the *v*_{i}/*v*_{t} should be <1 even if the *v*_{i} is measured in the lead where its value is the greatest. The Brugada criteria are not widely accepted as standard for WCT evaluation, thus, the superiority of the new to the Brugada algorithm demonstrated in this study does not necessarily imply that the new algorithm may be the best current method for WCT evaluation. However, Brugada *et al*.^{7} claimed that their algorithm had a better sensitivity and specificity than the traditional criteria without providing a true head-to-head comparison of the two methods in their study. Although other authors^{1,10,18} reported a lower sensitivity and specificity of the Brugada algorithm than those originally reported by Brugada *et al*.^{7} they still found the Brugada citeria useful and the claim that the Brugada criteria are superior to the traditional criteria was neither confirmed nor refuted in a study using a head-to-head comparison of the two methods.

## Conclusions

By using all published ECG criteria, the underlying cause of regular WCTs is still misdiagnosed in up to 10% of patients. It seems prudent to consider and treat all sustained, regular WCTs as VT unless the diagnosis of SVT can be definitely established, because it is far better to be wrong with a few cases of SVT treated as VT than the reverse situation, since treating a VT as SVT may result in potentially disastrous consequences (e.g. iv verapamil injection may cause severe hypotension and/or VT acceleration and ventricular fibrillation^{2,26,27}). The proposed new algorithm, which includes two new ECG criteria, may be useful to improve our diagnostic accuracy.

## Acknowledgements

We are indebted to Naomi Fineberg PhD, Júlia Singer and Elek Dinya for their valuable help in the statistical evaluation of the study.

**Conflict of interest**: none declared.

- © The European Society of Cardiology 2007. All rights reserved. For Permissions, please e-mail: journals.permissions@oxfordjournals.org